Multispectral ex vivo photoacoustic imaging of cutaneous melanoma for better selection of the excision margin

Dear Editor, Cutaneous melanoma is the most common malignant skin cancer and its incidence and mortality rates have been increasing over the last few decades.1 The standard treatment method for cutaneous melanoma is surgical excision with a certain margin from the visible boundary of the tumour region.2 It is extremely important to choose an appropriate excision margin to reduce the surgical operation time, cost and patients' discomfort. However, complete excision of the tumour, particularly lentigo maligna or other forms of lentiginous superficial spreading melanomas, is difficult due to invisible invasion of melanoma into the surrounding tissue. The selection of the excision margin is typically decided by the colour of the skin and the surgeons' experience. Thus, additional noninvasive imaging methods that can provide information outlining the melanoma and its surrounding tissue are essential for surgeons to determine the optimal excision margin. Here we report our initial pilot study to demonstrate photoacoustic (PA) imaging as an imaging method to provide additional information on the tumour boundary. PA imaging is a noninvasive biomedical imaging technique, which has gained great interest in biomedical research due to its various advantages including noninvasiveness, nonionizing radiation, cost‐efficiency and functional imaging capability. The principle of PA imaging is energy transformation from light to ultrasound (US) wave.3 When a short pulsed laser illuminates the target tissue, the absorbed light energy is released as a sound wave called a PA wave. Optically absorbing substances in biological tissue such as oxy‐ and deoxyhaemoglobin, melanin and lipid can give endogenous contrast for PA imaging.4 In addition, by tuning the excitation wavelength, PA imaging can selectively visualize the optical characteristics of biological tissue.5 In recent years, PA imaging systems have been investigated worldwide for clinical applications.6,7 However, the previous clinical PA systems are limited because of their single anatomical availability (e.g. breast only), clinically irrelevant US transducer (e.g. high‐frequency transducer), immobility or unmodifiable operation sequence. We have recently developed a clinical real‐time PA and US imaging system by combining a Food and Drug Administration‐cleared US machine and a wavelength‐tuneable mobile laser.8 Because the system has a programmable platform that enables modification of operational sequences, users can perform their specific research by programming a specific sequence of operations on the US machine. In addition, as the system is based on a universal US machine, PA images of any body part can be acquired, which can be of great advantage in delineating the boundary of cutaneous melanoma. To acquire PA data, we recruited a 57‐year‐old male patient for the study who had an asymptomatic, solitary and blackish pigmented patch on the left heel. The size of the lesion was 4 × 4 cm2. In a preliminary histopathological examination, we found a lentiginous proliferation of abnormal melanocytes at the dermoepidermal junction. Because our imaging system is based on a conventional clinical ultrasound imaging system that produces two‐dimensional cross‐sectional B‐mode images, it is difficult to acquire an in vivo three‐dimensional volume‐rendering image. Therefore, we acquired three‐dimensional volumetric images of an ex vivo sample by adding a simple scanning system (figure available on request). All of the patient recruiting and data acquisition procedures were performed in accordance with protocols approved by the institutional review boards of the Seoul St Mary's Hospital and Pohang University of Science and Technology. The melanoma lesion was excised with a margin of around 1 cm from the visible boundary (Fig. 1a). Immediately after excision, we acquired PA data with optical wavelengths of 680, 800 and 1064 nm. For each optical wavelength, we stored volumetric PA data by scanning a 128‐element linear array transducer in the elevational direction, with a scanning range of 75 mm and a scanning speed of 2 mm s−1. Therefore, the actual scanning time is 37.5 s. The scanning range was determined to include the entire resected lesion with an additional margin. We repeated scanning for all of the optical wavelengths, and the total data acquisition time was < 10 min. After imaging, we confirmed that histopathological examination revealed that the melanoma cells were localized in the epidermis, and the type of melanoma was acral lentiginous melanoma in situ (figure available on request). The tumour was free from the distal, proximal, medial and lateral margins at distances of 0.5, 0.5, 0.3 and 0.6 cm, respectively. (a) Photograph and (b) ex vivo photoacoustic (PA) image of excised melanoma tissue. The melanoma regions are represented by dark‐red to bright‐yellow colour (white triangles), and the marking‐pen regions are represented by dark‐green to bright‐green colour (white arrows). From the stored volumetric data, we generated overlaid PA and US B‐mode images at every scanning step (movie available on request). We also reconstructed the maximum‐amplitude projected PA images, and then combined them into one image (Fig. 1b). In the combined image, we can clearly distinguish the PA signals of melanoma and the marking pen regions using a spectral analysis technique: dark red to bright yellow for melanoma, and dark green to bright green for marking pen. The amplitude of the PA signal indicates the amount of the melanocytes in the local area. The PA signals in melanoma, indicated by the white triangles, were predominantly generated by optical wavelengths of 800 and 1064 nm. In contrast, the PA signals from the marking pen regions indicated by the white arrows were predominantly generated from the 680‐nm laser. The PA imaging indicated areas of melanoma that are not visible to the surgeon (yellow triangles in Fig. 1b). In addition, the measured thickness of the melanoma region (420 ± 320 μm) matched well with the histopathological result. We also generated three‐dimensional rendering images using the distinguished PA signals of melanoma and marking pen (movies available on request). All of the calculations were performed by parallel processing using a graphics processing unit, and the total image generation time was around 5 min, including beamforming, image generation, three‐dimensional rendering and maximum‐amplitude projection. In summary, we report the initial results of our pilot study of PA imaging of cutaneous melanoma. We acquired volumetric PA data of a resected tumour lesion from the left heel of a patient, then successfully delineated the tumour region using spectral analysis of PA signals. The lateral (1.20 ± 0.14 mm) and axial (205 ± 15 μm) resolutions of the system are not enough to detect single melanocytes, and the system may give false positive results because it is difficult to differentiate melanin of tumour melanocytes from melanin contained in keratinocytes and normal melanocytes. However, the resulting image can provide the general boundary of the melanoma, which can be helpful for better selection of the excision margin. Although in vivo analysis and larger‐scale studies are required to validate the technique's capability and feasibility, we believe that the results of our initial pilot study show great potential for PA imaging as a preoperative screening tool. This will provide additional information on the melanoma lesions to surgeons and aid them in performing effective surgery. Funding sources: this study was supported by the research funds from an IITP ICT Consilience Creative Program IITP‐2018‐2011‐1‐00783); National Research Foundation of Korea (NRF) grants (NRF‐2015R1C1A2A01054767, NRF‐2011‐0030075 and NRF‐2017M3C1A3037762) funded by the Ministry of Science, ICT and Future Planning; and the Korea Health Technology R&D Project of the Ministry of Health and Welfare (HI15C1817), funded by the Ministry of Health & Welfare, Republic of Korea. Conflicts of interest: none to declare. J.K. and Y.H.K. contributed equally to this work.